Predictive Modeling and Equation Formulation for 3D‐Printed Reinforced Concrete Beams

This study focuses on developing predictive models and design equations for 3D-printed reinforced concrete beams, utilizing machine learning to enhance the accuracy of structural analysis. The research aimed to create a novel equation for predicting bending moments, considering factors like tensile strength, compressive strength, and reinforcement patterns. Key goals included optimizing reinforcement placement and comparing regression models (gradient boosting, lasso, polynomial, stepwise, and random forest). Experiments tested four reinforcement patterns (honeycomb, 3D-honeycomb, grid, and triangle) in concrete beams, with optimization techniques identifying the best reinforcement position. Machine learning models were evaluated using R2, MAE, and RMSE metrics. Results showed gradient boosting as the most accurate (R2 = 0.997), with the optimal reinforcement placement at 7.41 mm from the beam’s bottom. SHapley Additive exPlanations (SHAP) analysis highlighted the reinforcement’s tensile strength as the most influential factor, while concrete compressive strength had a minor negative effect. The study provides a validated design equation for 3D-printed concrete beams, enhancing flexural performance and supporting sustainable construction. By combining machine learning with structural engineering, this work advances innovation in 3D-printed concrete, offering cost-effective and efficient solutions for civil engineering applications.